Monocular vision measurement error prediction and feedback control for a planar manipulator

Due to edge feature defects of markers, the visual measurements for the servo control of planar manipulators often exhibit systematically distributed errors even after conventional distortion correction and camera calibration. Then, the measurement errors of visual servo system have strongly limited the absolute positioning accuracy for the feedback control of the planar manipulators. In this paper, a monocular visual measurement error prediction and feedback control method is proposed for a planar manipulator. First, the monocular vision captures markers and extracts pixel information regarding the manipulator's endpoint, while the data from the laser tracker is converted into the same coordinate system to compare and compensate the visual measurement errors. Second, the visual measurement errors of the entire workspace are obtained using visual measurement error prediction based on radial basis function neural network spatial interpolation. The measurement accuracy of the visual measurement system was improved by 91.1%. Finally, the visual measurement error prediction-based feedback controls are conducted for the positioning of planar manipulator. The experiments demonstrate that the positioning accuracy of the planar manipulator is significantly improved via the applications of the proposed method. Its positioning accuracy can reach ±0.2 mm.